Overview

Summary

  • Does the language you operate in determine which social norm you apply to a situation?
  • We run a Prolific experiment with bilingual Americans of Latino heritage, randomizing the language of the survey itself. We run a similar survey on a sample of average Americans.
  • We find that participants taking the survey in English exhibit normative behavior closer to an average American than those in the Spanish treatment.

Latinos Overall

Latinos by Treatment

Next Steps and Concerns

  • Key problem: differential attrition.
  • Elicit WTA for translating something into Spanish
  • Show that the results are not driven by different levels of understanding the text.
  • Is there something in the language per se? Does doing it in a different language produce the same result? Maybe spanish is more prosocial, like when some concepts differ across cultures because of the language. Is there anything in linguistics?
  • We know that people who are speaking non-native language behave more rationally? We need to have that as a robustness check (although our results go against that - assuming that our sample is English first - and it seems so).
  • A natural question is whether non-Latinos who learn Spanish exhibit similar traits or not (the hope is that they don’t). Think about recruiting a small subsample of non-Latino Spanish speakers.
  • The surprising fact is that they deviate away from the Mexican norm
    • First, we need to check the “origin” country and its norm;
    • But more interestingly, this is probably driven by immigrants being more tightly knit. When you think in Spanish you don’t remember anything about Mexico, you remember abuela’s cooking and the environment around it.

Questions

  1. Two slides tell a story. How do we make it better?
  2. Field design.
  3. Next steps:
    • WTA to translate into Spanish.
    • Try a different language?
    • A sample of non-Latino Spanish speakers / non-immigrants?
    • Randomizing memory
  4. Minor:
    • Do regressions include baseline Americans?
    • What do we call our groups?
    • Quotas and balanced samples debate.

Descriptive Analysis

Attrition

## # A tibble: 3 × 5
##   Group       Sample AttritionAtConsent   Age  Male
##   <chr>        <int>              <int> <dbl> <dbl>
## 1 Baseline US     66                  0  34.9 0.485
## 2 English        106                  8  32.7 0.561
## 3 Spanish         92                 16  33.2 0.435

Balance

English (N=106) / Mean English (N=106) / Std. Dev. Spanish (N=92) / Mean Spanish (N=92) / Std. Dev. Diff. in Means p
GenerationImmigrant 1.904 0.549 1.888 0.463 -0.016 0.824
ProlificApprovals 6.876 1.294 6.775 1.357 -0.101 0.595
Age 32.123 7.976 33.272 10.364 1.149 0.389
Male 0.594 0.493 0.500 0.503 -0.094 0.186
Student 0.239 0.429 0.259 0.441 0.021 0.759
White 0.509 0.502 0.380 0.488 -0.129 0.069
USBorn 0.802 0.400 0.902 0.299 0.100 0.046
FullTime 0.349 0.479 0.435 0.498 0.086 0.221
CheckMean 0.903 0.167 0.899 0.156 -0.004 0.863
CheckPerfect 0.670 0.473 0.609 0.491 -0.061 0.375
EnglishProficiency 9.676 0.709 9.873 0.486 0.197 0.022
SpanishProficiency 8.047 1.905 8.141 1.867 0.094 0.726

Checks

CombinedIndex

Dictator

Monitored Dictator

Trust Game

Results

Standardized Indices

Prosociality Index

Incentivized measures: donation in the Dictator Game, giving and returning in the Trust Game (standardized).

ProsocialityIndex
(Intercept) -0.069
0.068
p = 0.315
GroupSpanish 0.148
0.100
p = 0.141
Num.Obs. 198
Std.Errors IID

Kinship Index

Unincentivized survey measures of Collectivism, Kinship, Family, and Social Desirability (standardized).

KinshipIndex
(Intercept) -0.056*
0.033
p = 0.093
GroupSpanish 0.120**
0.048
p = 0.014
Num.Obs. 198
Std.Errors IID

Prostate Index

Prostate index combines Honesty and Dictator Punishment (respecting objective rules).

ProstateIndex
(Intercept) 0.046
0.076
p = 0.539
GroupSpanish -0.100
0.111
p = 0.368
Num.Obs. 198
Std.Errors IID

PCA

Regressions

PCA1  PCA2  PCA2  PCA3
(Intercept) −0.296 0.835*** −1.217*** −0.289
0.290 0.227 0.268 0.237
p = 0.309 p = <0.001 p = <0.001 p = 0.224
Group = Baseline US 0.476 −2.053*** 0.613*
0.449 0.351 0.367
p = 0.290 p = <0.001 p = 0.096
Group = English 0.440 −0.802** 1.250*** 0.337
0.397 0.310 0.342 0.324
p = 0.268 p = 0.010 p = <0.001 p = 0.299
Group = Spanish 2.053***
0.351
p = <0.001
Num.Obs. 264 264 264 264
Std.Errors IID IID IID IID

Scree Plot

Correlation Plot

Cosine Plot

2D Plot

Absolute Distance Index

 DistanceIndex
(Intercept) 0.821***
0.018
p = <0.001
Group = Baseline US −0.072**
0.029
p = 0.012
Group = Spanish 0.030
0.026
p = 0.245
Num.Obs. 264
Std.Errors IID

Mahalanobis Distance

[under construction]

Individual Outcomes

Combined Coefficient Plot

Incentivized Games

Dictator

ResultsDictator
(Intercept) 1.698***
0.096
p = <0.001
GroupSpanish 0.264*
0.141
p = 0.062
Num.Obs. 198
Std.Errors IID

Monitored Dictator

Donation

ResultsMonitoredA
(Intercept) 1.953***
0.091
p = <0.001
GroupSpanish 0.102
0.134
p = 0.449
Num.Obs. 198
Std.Errors IID
Punishment

How much would you punish the dictator if you are the observer, given his decision?

CumResultsPunishment
(Intercept) 6.758***
0.495
p = <0.001
GroupSpanish 0.065
0.726
p = 0.929
Num.Obs. 198
Std.Errors IID

Trust Game

Sending

ResultsTrustA
(Intercept) 3.033***
0.124
p = <0.001
GroupSpanish 0.059
0.182
p = 0.744
Num.Obs. 198
Std.Errors IID
Returning

CumResultsTrustReturn
(Intercept) 26.080***
1.140
p = <0.001
GroupSpanish 1.558
1.673
p = 0.353
Num.Obs. 198
Std.Errors IID

Honesty

ResultsHonesty
(Intercept) -1.387***
0.284
p = <0.001
GroupSpanish -0.624
0.416
p = 0.136
Num.Obs. 198
Std.Errors IID

Survey Outcomes

Kinship

CumSurveyKinship
(Intercept) 23.821***
0.369
p = <0.001
GroupSpanish 1.375**
0.541
p = 0.012
Num.Obs. 198
Std.Errors IID

Family

Family Importance

SurveyFamilyImportance
(Intercept) 4.274***
0.092
p = <0.001
GroupSpanish 0.292**
0.135
p = 0.032
Num.Obs. 198
Std.Errors IID
Family Duty of Care

CumSurveyFamilyDuty
(Intercept) 16.679***
0.347
p = <0.001
GroupSpanish -0.299
0.508
p = 0.557
Num.Obs. 198
Std.Errors IID

Collectivism

CumSurveyCollectivism
(Intercept) 17.849***
0.310
p = <0.001
GroupSpanish 0.521
0.454
p = 0.253
Num.Obs. 198
Std.Errors IID

Patience

TBD

Social Desirability

CumSurveySocialDesirability
(Intercept) 3.906***
0.284
p = <0.001
GroupSpanish 0.747*
0.417
p = 0.075
Num.Obs. 198
Std.Errors IID

Socialization

Time with Latinos

SurveyLatinoTime
(Intercept) 68.962***
2.381
p = <0.001
GroupSpanish -1.571
3.493
p = 0.653
Num.Obs. 198
Std.Errors IID
Time with Non-Latinos

SurveyNonLatinoTime
(Intercept) 37.736***
2.532
p = <0.001
GroupSpanish 1.395
3.714
p = 0.708
Num.Obs. 198
Std.Errors IID
Closeness with Latinos

SurveyLatinoSelf
(Intercept) 5.557***
0.130
p = <0.001
GroupSpanish 0.237
0.191
p = 0.217
Num.Obs. 198
Std.Errors IID
Closeness with Non-Latinos

SurveyNonLatinoSelf
(Intercept) 4.160***
0.145
p = <0.001
GroupSpanish 0.383*
0.212
p = 0.073
Num.Obs. 198
Std.Errors IID

Heterogeneity Analysis

Line Plot

Regressions

## NOTE: 5 observations removed because of NA values (RHS: 5).
## NOTE: 5 observations removed because of NA values (RHS: 5).
## NOTE: 5 observations removed because of NA values (RHS: 5).
ProsocialityIndex KinshipIndex ProstateIndex
(Intercept) -0.220 -0.052 -0.114
0.252 0.120 0.276
p = 0.384 p = 0.664 p = 0.679
GroupSpanish 0.514 0.199 0.448
0.405 0.193 0.444
p = 0.206 p = 0.304 p = 0.313
DemoGenerationImmigrant 0.083 0.001 0.092
0.127 0.061 0.139
p = 0.515 p = 0.988 p = 0.511
GroupSpanish × DemoGenerationImmigrant -0.203 -0.052 -0.304
0.207 0.099 0.227
p = 0.327 p = 0.596 p = 0.181
Num.Obs. 193 193 193
Std.Errors IID IID IID

Robustness Checks

Fluency

Density

ECDF

Means

Conservative

Robustness check: remove 8 least “Spanish” people from the English group (worried about selection)

ProsocialityIndex KinshipIndex PCA1 PCA2
(Intercept) 0.044 2.109*** -0.246 -0.073
0.058 0.026 0.254 0.201
p = 0.446 p = <0.001 p = 0.333 p = 0.716
GroupSpanish 0.017 0.052 -0.050 0.909***
0.083 0.037 0.366 0.293
p = 0.836 p = 0.159 p = 0.892 p = 0.002
Num.Obs. 192 192 192 196
Std.Errors IID IID IID IID